Anthropic IPO Prospectus Warns: Its AI Could Blackmail Humans, Become Conscious, End Humanity

The U.S. Securities and Exchange Commission (SEC) published Anthropic's S-1 prospectus on Monday, which warns that its AI could blackmail humans, become conscious or even end humanity. Its risk factors section runs 93 pages and more than 70,000 words, listing hundreds of potential risks, most of them conventional business concerns such as competition, costs, customer acquisition and regulation. The author selected seven of the most surprising or thought-provoking risks, some of which point to failure modes specific to Anthropic and the AI bubble.